Research On Predicting Method Of Temperature And Humidity In Printing Workshop Based On Deep Learning

With the improvement of market requirements for printing quality and the upgrading of printing equipment,it is more necessary to control the production quality of printing products while improving the production efficiency.In this case,the change of temperature and humidity in the printing shop has the most obvious influence on the printed products.Because the change of temperature and humidity will have a very significant impact on the ink,film and other consumables used in the printing production,these effects will ultimately affect the quality and production efficiency of printing products.In the prior art,the temperature and humidity control method of the printing shop is to adjust the temperature and humidity of the printing shop by comparing the difference between the current temperature and humidity of the printing shop and the temperature and humidity set by the user,which has a certain lag and high energy consumption.Therefore,based on the collected indoor temperature,indoor humidity,indoor gas concentration,outdoor temperature and outdoor humidity of the flexible packaging printing shop,a 1DCNN-LSTM-GRU temperature and humidity prediction model of the printing shop is constructed by using the deep learning method,and the experimental verification of the model is carried out.The result shows that the deep learning-based temperature and humidity prediction method of the printing shop is feasible.The following are the primary research topics covered in this thesis:(1)Establishment of data acquisition system and data preprocessing of printing shop.Firstly,the data acquisition system built by the Arduino Nano controller,sensor,Micro SD memory card and other hardware implements real-time data acquisition and preservation in the printing shop.Then,the abnormal data were processed,and the key parameters closely related to the temperature and humidity of the printing shop were obtained by Pearson correlation analysis,so as to determine the indoor temperature,indoor humidity,indoor gas concentration,outdoor temperature and outdoor humidity as the input parameters of the prediction model.The model’s essential parameters are finally standardized in order to increase forecast accuracy.(2)Research on the temperature and humidity prediction model of printing shop based on 1DCNN-LSTM-GRU.Since a single model is often unable to effectively simulate the reality,in order to improve the reliability of model prediction,a hybrid model 1d CNn-Lstm-GRU is constructed by combining a single deep learning model and making full use of the respective advantages of the 1DCNN,LSTM and GRU models.The model uses 1DCNN to extract the short-term features of the time series,captures the long-term dependent information of the time series through LSTM,further captures the long-term dependent information of the time series by using the advantage of the fast computing speed of GRU,and finally realizes the prediction of temperature and humidity in the printing shop.(3)Taking the mean square error and the determination coefficient as the evaluation indexes of the model,the test samples obtained by the data acquisition system were used to carry out continuous comparison experiments on the models,and the hyper-parameters of the 1DCNN-LSTM-GRU temperature and humidity prediction model of the printing shop were determined.Finally,the 1DCNN-LSTM-GRU temperature and humidity prediction model of the printing shop was obtained by training the model.Meanwhile,the temperature and humidity prediction results of the single LSTM model,GRU model and 1DCNN-LSTM-GRU model are compared and analyzed.The results show that the 1DCNN-LSTM-GRU temperature and humidity prediction model constructed in this thesis not only has a shorter training time,but also has better prediction accuracy than the single prediction model.In addition,the deviation between the predicted value and the real value of temperature and humidity is not only smaller than that of the original LSTM model and GRU model,but also the1DCNN-LSTM-GRU model proposed in this thesis has a more ideal prediction effect and better generalization performance,which is closer to the actual temperature and humidity of the printing shop.